SkillSafeAgent workflow & memorySome setup

agentic-ai-systems

ThibautMelen · ThibautMelen/agentic-ai-systems

This resource explains agentic AI systems with runnable, CI-checked patterns, covering foundations, workflows, and implementation details for building multi-agent architectures.

Agentic AI explained with chickens 🐔 every pattern a runnable, CI-checked file.

315 stars35License: MITUpdated 3 months ago0 opens

Preview

▍

What you can do with it

Learn various agentic workflow patterns like chaining, routing, and orchestration.
Implement subagents, commands, skills, and hooks in your own agentic system.
Apply best practices for context engineering and multi-agent debate in AI applications.

Safety check

Safe

No risky patterns found in the files we scanned.

Automated scan of the repo files. Always review code you run.

From the README

Agentic AI Systems 🐔

Agentic systems explained with chickens — because AI orchestration doesn't have to be scary

Main Agent spawns Subagents like a hen with her chicks 🐔🪺🐦

🦄 Foundations • ⚙️ Workflows • 🐔 Autonomous • 📜 As Code • 🛠️ Implementation • 🗺️ Guides

Overview

%%{init: {'theme': 'base', 'themeVariables': { 'primaryColor': '#8b5cf6', 'primaryTextColor': '#fff', 'primaryBorderColor': '#7c3aed', 'lineColor': '#a78bfa', 'secondaryColor': '#ec4899', 'tertiaryColor': '#6366f1', 'noteTextColor': '#fff', 'noteBkgColor': '#8b5cf6', 'textColor': '#fff' }}}%%
mindmap
  root((🐔 Agentic Systems))
    🦄 Foundations
      Augmented LLM
      What Changed 2026
    ⚙️ Workflows
      🏎️ Baseline
      ⛓️ Chaining
      🚦 Routing
      🛤️ Parallel
      🦑 Orchestrator
      🩻 Evaluator
    🐔 Autonomous Agent
      The Alternative
      Multi-Window
    🛠️ Implementation
      🐦 Subagent
      🦴 Command
      📚 Skill
      🪝 Hook
    📜 Patterns as Code
      Check before
      Trace after

[!TIP] New in the 2026 edition — 🗞️ What Changed 2025-2026 (the dated map: context engineering, the multi-agent debate, the AGENTS.md · SKILL.md · MCP stack, harness engineering) and 📜 Patterns as Code (every pattern as a runnable file — checked in CI on every push, no API key needed).

⬆️ real capture: a repeated prompt graduates into a checked, runnable file — every pattern in this repo works this way · and three of them run a real campaign daily 🛰️

🗺️ Navigation

🦄 Foundations

The building block for everything


🦄 Augmented LLM	LLM + Retrieval + Tools + Memory
🗞️ What Changed 2025-2026	Context engineering · skills/MCP stack · harness era

⚙️ Workflows

Predefined orchestration — code controls the flow

#	Workflow	Use When
0	🏎️ Baseline	Simple, 1-step task
1	⛓️ Prompt Chaining	Sequential steps
2	🚦 Routing	Classify & dispatch
3	🛤️ Parallelization	Independent tasks
4	🦑 Orchestrator	Expert delegation
5	🩻 Evaluator	Quality iteration

🐔 Autonomous Agent

Dynamic autonomy — LLM controls the flow

Pattern	Use When
🐔 Autonomous Agent	Open-ended problems
🖥️ Multi-Window	Cross-session state (variant)

🛠️ Implementation

Claude Code components & architecture

Component	Location
🐦 Subagent	`.claude/agents/*.md`
🦴 Command	`.claude/commands/*.md`
📚 Skill	`.claude/skills/*/SKILL.md`
🪝 Hook	`.claude/settings.json`

🗺️ Guides & 📖 Reference

Resource	Description
Selection Guide	Choose the right pattern
📜 Patterns as Code	Every pattern as a runnable, checkable file
Use Cases	6 validated examples
Glossary	A-Z definitions
Visual Standards	Colors & emojis

Quick Decision

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart LR
    START((🎯 Task)) --> DEST{Destructive?}
    DEST -->|Yes| WIZ[🧙 Wizard]
    DEST -->|No| COMP{Complex?}
    COMP -->|No| BASE[🏎️ Baseline]
    COMP -->|Yes| PRED{Predictablesteps?}
    PRED -->|Yes| WORK{Needspecialists?}
    PRED -->|No| AGENT[🐔 Autonomous]
    WORK -->|No| CHAIN[⛓️ Chain]
    WORK -->|Yes| ORCH[🦑 Orchestrator]

    classDef default fill:#f8fafc,stroke:#64748b,stroke-width:1px,color:#1e293b
    classDef decision fill:#fef3c7,stroke:#f59e0b,stroke-width:2px,color:#92400e
    classDef baseline fill:#64748b,stroke:#475569,stroke-width:2px,color:#ffffff
    classDef wizard fill:#14b8a6,stroke:#0d9488,stroke-width:2px,color:#ffffff
    classDef workflow fill:#8b5cf6,stroke:#7c3aed,stroke-width:2px,color:#ffffff
    classDef agent fill:#ec4899,stroke:#db2777,stroke-width:2px,color:#ffffff

    START:::decision
    DEST:::decision
    COMP:::decision
    PRED:::decision
    WORK:::decision
    BASE:::baseline
    WIZ:::wizard
    CHAIN:::workflow
    ORCH:::workflow
    AGENT:::agent

Situation	→ Use
Simple task (1 step)	🏎️ Baseline
Sequential (2-4 steps)	⛓️ Prompt Chaining
Categorize inputs	🚦 Routing
Independent subtasks	🛤️ Parallelization
Multiple specialists	🦑 Orchestrator-Workers
Quality iteration	🩻 Evaluator-Optimizer
Open-ended / unknown steps	🐔 Autonomous Agent
Destructive operations	🧙 Wizard
Long-running (>10 min)	🖥️ Multi-Window Context

Anthropic Taxonomy

%%{init: {'theme': 'base', 'themeVariables': {'lineColor': '#64748b'}}}%%
flowchart LR
    subgraph WORKFLOWS["⚙️ WORKFLOWS"]
        direction TB
        W1[🏎️ Baseline]
        W2[⛓️ Prompt Chaining]
        W3[🚦 Routing]
        W4

View source on GitHub